On 10/18/24 01:11, Xu Yilun wrote:
Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding. On Mon, Oct 07, 2024 at 03:01:27PM -0700, David Zhang wrote:From: Yidong Zhang <yidong.zhang@xxxxxxx> Add queue based communication between host driver and firmware on the card. The remote queue (rm) can send/receive messages and providingAbbrevate 'remote queue' to 'rm', or 'remote' to 'rm'. I see you use rm_queue in the code.
I will fix it.
firmware downloading services. Co-developed-by: Nishad Saraf <nishads@xxxxxxx> Signed-off-by: Nishad Saraf <nishads@xxxxxxx> Co-developed-by: Prapul Krishnamurthy <prapulk@xxxxxxx> Signed-off-by: Prapul Krishnamurthy <prapulk@xxxxxxx> Signed-off-by: Yidong Zhang <yidong.zhang@xxxxxxx> --- drivers/fpga/amd/Makefile | 6 +- drivers/fpga/amd/vmgmt-rm-queue.c | 38 +++ drivers/fpga/amd/vmgmt-rm-queue.h | 15 + drivers/fpga/amd/vmgmt-rm.c | 543 ++++++++++++++++++++++++++++++ drivers/fpga/amd/vmgmt-rm.h | 222 ++++++++++++ drivers/fpga/amd/vmgmt.c | 305 ++++++++++++++++- drivers/fpga/amd/vmgmt.h | 7 +- 7 files changed, 1130 insertions(+), 6 deletions(-) create mode 100644 drivers/fpga/amd/vmgmt-rm-queue.c create mode 100644 drivers/fpga/amd/vmgmt-rm-queue.h create mode 100644 drivers/fpga/amd/vmgmt-rm.c create mode 100644 drivers/fpga/amd/vmgmt-rm.h diff --git a/drivers/fpga/amd/Makefile b/drivers/fpga/amd/Makefile index 3e4c6dd3b787..97cfff6be204 100644 --- a/drivers/fpga/amd/Makefile +++ b/drivers/fpga/amd/Makefile @@ -2,5 +2,7 @@ obj-$(CONFIG_AMD_VERSAL_MGMT) += amd-vmgmt.o -amd-vmgmt-$(CONFIG_AMD_VERSAL_MGMT) := vmgmt.o \ - vmgmt-comms.o +amd-vmgmt-$(CONFIG_AMD_VERSAL_MGMT) := vmgmt.o \ + vmgmt-comms.o \ + vmgmt-rm.o \ + vmgmt-rm-queue.o diff --git a/drivers/fpga/amd/vmgmt-rm-queue.c b/drivers/fpga/amd/vmgmt-rm-queue.c new file mode 100644 index 000000000000..fe805373ea32 --- /dev/null +++ b/drivers/fpga/amd/vmgmt-rm-queue.c @@ -0,0 +1,38 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Driver for Versal PCIe device + * + * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved. + */ + +#include <linux/bitfield.h> +#include <linux/completion.h> +#include <linux/err.h> +#include <linux/firmware.h> +#include <linux/idr.h> +#include <linux/jiffies.h> +#include <linux/list.h> +#include <linux/mutex.h> +#include <linux/pci.h> +#include <linux/semaphore.h> +#include <linux/timer.h> +#include <linux/uuid.h> +#include <linux/workqueue.h> + +#include "vmgmt.h" +#include "vmgmt-rm.h" +#include "vmgmt-rm-queue.h" + +int rm_queue_send_cmd(struct rm_cmd *cmd, unsigned long timeout) +{ + return 0; +} + +void rm_queue_fini(struct rm_device *rdev) +{ +} + +int rm_queue_init(struct rm_device *rdev) +{ + return 0; +} diff --git a/drivers/fpga/amd/vmgmt-rm-queue.h b/drivers/fpga/amd/vmgmt-rm-queue.h new file mode 100644 index 000000000000..6fd0e0026a13 --- /dev/null +++ b/drivers/fpga/amd/vmgmt-rm-queue.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Driver for Versal PCIe device + * + * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved. + */ + +#ifndef __VMGMT_RM_QUEUE_H +#define __VMGMT_RM_QUEUE_H + +int rm_queue_init(struct rm_device *rdev); +void rm_queue_fini(struct rm_device *rdev); +int rm_queue_send_cmd(struct rm_cmd *cmd, unsigned long timeout); + +#endif /* __VMGMT_RM_QUEUE_H */ diff --git a/drivers/fpga/amd/vmgmt-rm.c b/drivers/fpga/amd/vmgmt-rm.c new file mode 100644 index 000000000000..856d5af52c8d --- /dev/null +++ b/drivers/fpga/amd/vmgmt-rm.c @@ -0,0 +1,543 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Driver for Versal PCIe device + * + * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved. + */ + +#include <linux/bitfield.h> +#include <linux/completion.h> +#include <linux/err.h> +#include <linux/firmware.h> +#include <linux/idr.h> +#include <linux/jiffies.h> +#include <linux/list.h> +#include <linux/mutex.h> +#include <linux/pci.h> +#include <linux/semaphore.h> +#include <linux/timer.h> +#include <linux/uuid.h> +#include <linux/workqueue.h> + +#include "vmgmt.h" +#include "vmgmt-rm.h" +#include "vmgmt-rm-queue.h" + +static DEFINE_IDA(rm_cmd_ids); + +static const struct regmap_config rm_shmem_regmap_config = { + .name = "rm_shmem_config", + .reg_bits = 32, + .reg_stride = 4, + .val_bits = 32, +}; + +static const struct regmap_config rm_io_regmap_config = { + .name = "rm_io_config", + .reg_bits = 32, + .reg_stride = 4, + .val_bits = 32, +}; + +static void rm_uninstall_health_monitor(struct rm_device *rdev); + +static inline struct rm_device *to_rdev_health_monitor(struct work_struct *w) +{ + return container_of(w, struct rm_device, health_monitor); +} + +static inline struct rm_device *to_rdev_health_timer(struct timer_list *t) +{ + return container_of(t, struct rm_device, health_timer); +} + +static inline int rm_shmem_read(struct rm_device *rdev, u32 offset, u32 *value) +{ + return regmap_read(rdev->shmem_regmap, offset, value); +} + +static inline int rm_shmem_bulk_read(struct rm_device *rdev, u32 offset, + u32 *value, u32 size) +{ + return regmap_bulk_read(rdev->shmem_regmap, offset, value, + DIV_ROUND_UP(size, 4)); +} + +static inline int rm_shmem_bulk_write(struct rm_device *rdev, u32 offset, + u32 *value, u32 size) +{ + return regmap_bulk_write(rdev->shmem_regmap, offset, value, + DIV_ROUND_UP(size, 4)); +} + +void rm_queue_destory_cmd(struct rm_cmd *cmd) +{ + ida_free(&rm_cmd_ids, cmd->sq_msg.hdr.id); + kfree(cmd); +} + +int rm_queue_copy_response(struct rm_cmd *cmd, void *buffer, ssize_t len) +{ + struct rm_cmd_cq_log_page *result = &cmd->cq_msg.data.page; + u64 off = cmd->sq_msg.data.page.address; + + if (!result->len || len < result->len) { + vmgmt_err(cmd->rdev->vdev, "Invalid response or buffer size"); + return -EINVAL; + } + + return rm_shmem_bulk_read(cmd->rdev, off, (u32 *)buffer, result->len); +} + +static void rm_queue_payload_fini(struct rm_cmd *cmd) +{ + up(&cmd->rdev->cq.data_lock); +} + +static int rm_queue_payload_init(struct rm_cmd *cmd, + enum rm_cmd_log_page_type type) +{ + struct rm_device *rdev = cmd->rdev; + int ret; + + ret = down_interruptible(&rdev->cq.data_lock); + if (ret) + return ret; + + cmd->sq_msg.data.page.address = rdev->cq.data_offset; + cmd->sq_msg.data.page.size = rdev->cq.data_size; + cmd->sq_msg.data.page.reserved1 = 0; + cmd->sq_msg.data.page.type = FIELD_PREP(RM_CMD_LOG_PAGE_TYPE_MASK, + type); + return 0; +} + +void rm_queue_data_fini(struct rm_cmd *cmd) +{ + up(&cmd->rdev->sq.data_lock); +} + +int rm_queue_data_init(struct rm_cmd *cmd, const char *buffer, ssize_t size) +{ + struct rm_device *rdev = cmd->rdev; + int ret; + + if (!size || size > rdev->sq.data_size) { + vmgmt_err(rdev->vdev, "Unsupported file size"); + return -ENOMEM; + } + + ret = down_interruptible(&rdev->sq.data_lock); + if (ret) + return ret; + + ret = rm_shmem_bulk_write(cmd->rdev, rdev->sq.data_offset, + (u32 *)buffer, size); + if (ret) { + vmgmt_err(rdev->vdev, "Failed to copy binary to SQ buffer"); + up(&cmd->rdev->sq.data_lock); + return ret; + } + + cmd->sq_msg.data.bin.address = rdev->sq.data_offset; + cmd->sq_msg.data.bin.size = size; + return 0; +} + +int rm_queue_create_cmd(struct rm_device *rdev, enum rm_queue_opcode opcode, + struct rm_cmd **cmd_ptr) +{ + struct rm_cmd *cmd = NULL; + int ret, id; + u16 size; + + if (rdev->firewall_tripped) + return -ENODEV; + + cmd = kzalloc(sizeof(*cmd), GFP_KERNEL); + if (!cmd) + return -ENOMEM; + cmd->rdev = rdev; + + switch (opcode) { + case RM_QUEUE_OP_LOAD_XCLBIN: + fallthrough; + case RM_QUEUE_OP_LOAD_FW: + fallthrough; + case RM_QUEUE_OP_LOAD_APU_FW: + size = sizeof(struct rm_cmd_sq_bin); + break; + case RM_QUEUE_OP_GET_LOG_PAGE: + size = sizeof(struct rm_cmd_sq_log_page); + break; + case RM_QUEUE_OP_IDENTIFY: + size = 0; + break; + case RM_QUEUE_OP_VMR_CONTROL: + size = sizeof(struct rm_cmd_sq_ctrl); + break; + default: + vmgmt_err(rdev->vdev, "Invalid cmd opcode %d", opcode); + ret = -EINVAL; + goto error; + }; + + cmd->opcode = opcode; + cmd->sq_msg.hdr.opcode = FIELD_PREP(RM_CMD_SQ_HDR_OPS_MSK, opcode); + cmd->sq_msg.hdr.msg_size = FIELD_PREP(RM_CMD_SQ_HDR_SIZE_MSK, size); + + id = ida_alloc_range(&rm_cmd_ids, RM_CMD_ID_MIN, RM_CMD_ID_MAX, GFP_KERNEL); + if (id < 0) { + vmgmt_err(rdev->vdev, "Failed to alloc cmd ID: %d", id); + ret = id; + goto error; + } + cmd->sq_msg.hdr.id = id; + + init_completion(&cmd->executed); + + *cmd_ptr = cmd; + return 0; +error: + kfree(cmd); + return ret; +} + +static int rm_queue_verify(struct rm_device *rdev) +{ + struct vmgmt_device *vdev = rdev->vdev; + struct rm_cmd_cq_identify *result; + struct rm_cmd *cmd; + u32 major, minor; + int ret; + + ret = rm_queue_create_cmd(rdev, RM_QUEUE_OP_IDENTIFY, &cmd); + if (ret) + return ret; + + ret = rm_queue_send_cmd(cmd, RM_CMD_WAIT_CONFIG_TIMEOUT); + if (ret) + goto error; + + result = &cmd->cq_msg.data.identify; + major = result->major; + minor = result->minor; + vmgmt_dbg(vdev, "VMR version %d.%d", major, minor); + if (!major) { + vmgmt_err(vdev, "VMR version is unsupported"); + ret = -EOPNOTSUPP; + } + +error: + rm_queue_destory_cmd(cmd); + return ret; +} + +static int rm_check_apu_status(struct rm_device *rdev, bool *status) +{ + struct rm_cmd_cq_control *result; + struct rm_cmd *cmd; + int ret; + + ret = rm_queue_create_cmd(rdev, RM_QUEUE_OP_VMR_CONTROL, &cmd); + if (ret) + return ret; + + ret = rm_queue_send_cmd(cmd, RM_CMD_WAIT_CONFIG_TIMEOUT); + if (ret) + goto error; + + result = &cmd->cq_msg.data.ctrl; + *status = FIELD_GET(RM_CMD_VMR_CONTROL_PS_MASK, result->status); + + rm_queue_destory_cmd(cmd); + return 0; + +error: + rm_queue_destory_cmd(cmd); + return ret; +} + +static int rm_download_apu_fw(struct rm_device *rdev, char *data, ssize_t size) +{ + struct rm_cmd *cmd; + int ret; + + ret = rm_queue_create_cmd(rdev, RM_QUEUE_OP_LOAD_APU_FW, &cmd); + if (ret) + return ret; + + ret = rm_queue_data_init(cmd, data, size); + if (ret) + goto done; + + ret = rm_queue_send_cmd(cmd, RM_CMD_WAIT_DOWNLOAD_TIMEOUT); + +done: + rm_queue_destory_cmd(cmd); + return ret; +} + +int rm_boot_apu(struct rm_device *rdev) +{ + char *bin = "xilinx/xrt-versal-apu.xsabin";So apu fw never changes, is it?
The APU FW does not change when the card is in service. It works with the base shell FW together. The location of the "*bin" does not change too. The binary xsabin can be changed by installing a deb pkg
+ const struct firmware *fw = NULL; + bool status; + int ret; + + ret = rm_check_apu_status(rdev, &status); + if (ret) { + vmgmt_err(rdev->vdev, "Failed to get APU status"); + return ret; + } + + if (status) { + vmgmt_dbg(rdev->vdev, "APU online. Skipping APU FW download"); + return 0; + } + + ret = request_firmware(&fw, bin, &rdev->vdev->pdev->dev); + if (ret) { + vmgmt_warn(rdev->vdev, "Request APU FW %s failed %d", bin, ret); + return ret; + } + + vmgmt_dbg(rdev->vdev, "Starting... APU FW download"); + ret = rm_download_apu_fw(rdev, (char *)fw->data, fw->size); + vmgmt_dbg(rdev->vdev, "Finished... APU FW download %d", ret); + + if (ret) + vmgmt_err(rdev->vdev, "Failed to download APU FW, ret:%d", ret); + + release_firmware(fw); + + return ret; +} + +static void rm_check_health(struct work_struct *w) +{ + struct rm_device *rdev = to_rdev_health_monitor(w); + ssize_t len = PAGE_SIZE; + char *buffer = NULL; + struct rm_cmd *cmd; + int ret; + + buffer = vzalloc(len); + if (!buffer) + return; + + ret = rm_queue_create_cmd(rdev, RM_QUEUE_OP_GET_LOG_PAGE, &cmd); + if (ret) + return;Memory Leak!
I will fix it.
+ + ret = rm_queue_payload_init(cmd, RM_CMD_LOG_PAGE_AXI_TRIP_STATUS); + if (ret) + goto error; + + ret = rm_queue_send_cmd(cmd, RM_CMD_WAIT_CONFIG_TIMEOUT); + if (ret == -ETIME || ret == -EINVAL) + goto payload_fini; + + if (cmd->cq_msg.data.page.len) {If no data, the heath is good?
If not data and ret value is not ZERO, the health is not good.The current FW guarrents that there will be always some data in this situation. I can add code like:
if (ret) {
if (has data)
printk(data);
else
printk(readable info based on ret value);
}
+ ret = rm_queue_copy_response(cmd, buffer, len); + if (ret) + goto payload_fini; + + vmgmt_err(rdev->vdev, "%s", buffer);Any concern reading out of bound. If the buffer is only used in this block, put its allocation/free here.
Yes, I will move the buffer here.
+ rdev->firewall_tripped = 1; + } + + vfree(buffer); + + rm_queue_payload_fini(cmd); + rm_queue_destory_cmd(cmd); + + return; + +payload_fini: + rm_queue_payload_fini(cmd); +error: + rm_queue_destory_cmd(cmd); + vfree(buffer); +} + +static void rm_sched_health_check(struct timer_list *t) +{ + struct rm_device *rdev = to_rdev_health_timer(t); + + if (rdev->firewall_tripped) { + vmgmt_err(rdev->vdev, "Firewall tripped, health check paused. Please reset card"); + return; + } + /* Schedule a work in the general workqueue */ + schedule_work(&rdev->health_monitor); + /* Periodic timer */ + mod_timer(&rdev->health_timer, jiffies + RM_HEALTH_CHECK_TIMER);Again, is it necessary we endless poll. Coundn't we check on cmd create?
The remote queue hardware only support polling mode.
+} + +static void rm_uninstall_health_monitor(struct rm_device *rdev) +{ + del_timer_sync(&rdev->health_timer); + cancel_work_sync(&rdev->health_monitor); +} + +static void rm_install_health_monitor(struct rm_device *rdev) +{ + INIT_WORK(&rdev->health_monitor, &rm_check_health); + timer_setup(&rdev->health_timer, &rm_sched_health_check, 0); + mod_timer(&rdev->health_timer, jiffies + RM_HEALTH_CHECK_TIMER); +} + +void vmgmt_rm_fini(struct rm_device *rdev) +{ + rm_uninstall_health_monitor(rdev); + rm_queue_fini(rdev); +} + +struct rm_device *vmgmt_rm_init(struct vmgmt_device *vdev) +{ + struct rm_header *header; + struct rm_device *rdev; + u32 status; + int ret; + + rdev = devm_kzalloc(&vdev->pdev->dev, sizeof(*rdev), GFP_KERNEL); + if (!rdev) + return ERR_PTR(-ENOMEM); + + rdev->vdev = vdev; + header = &rdev->rm_metadata; + + rdev->shmem_regmap = devm_regmap_init_mmio(&vdev->pdev->dev, + vdev->tbl + RM_PCI_SHMEM_BAR_OFF, + &rm_shmem_regmap_config); + if (IS_ERR(rdev->shmem_regmap)) { + vmgmt_err(vdev, "Failed to init RM shared memory regmap"); + return ERR_CAST(rdev->shmem_regmap); + } + + ret = rm_shmem_bulk_read(rdev, RM_HDR_OFF, (u32 *)header, + sizeof(*header)); + if (ret) { + vmgmt_err(vdev, "Failed to read RM shared mem, ret %d", ret); + ret = -ENODEV; + goto err; + } + + if (header->magic != RM_HDR_MAGIC_NUM) { + vmgmt_err(vdev, "Invalid RM header 0x%x", header->magic); + ret = -ENODEV; + goto err; + } + + ret = rm_shmem_read(rdev, header->status_off, &status); + if (ret) { + vmgmt_err(vdev, "Failed to read RM shared mem, ret %d", ret); + ret = -ENODEV; + goto err; + } + + if (!status) { + vmgmt_err(vdev, "RM status %d is not ready", status); + ret = -ENODEV; + goto err; + } + + rdev->queue_buffer_size = header->data_end - header->data_start + 1; + rdev->queue_buffer_start = header->data_start; + rdev->queue_base = header->queue_base; + + rdev->io_regmap = devm_regmap_init_mmio(&vdev->pdev->dev, + vdev->tbl + RM_PCI_IO_BAR_OFF, + &rm_io_regmap_config); + if (IS_ERR(rdev->io_regmap)) { + vmgmt_err(vdev, "Failed to init RM IO regmap"); + ret = PTR_ERR(rdev->io_regmap); + goto err; + } + + ret = rm_queue_init(rdev); + if (ret) { + vmgmt_err(vdev, "Failed to init cmd queue, ret %d", ret); + ret = -ENODEV; + goto err; + } + + ret = rm_queue_verify(rdev); + if (ret) { + vmgmt_err(vdev, "Failed to verify cmd queue, ret %d", ret); + ret = -ENODEV; + goto queue_fini; + } + + ret = rm_boot_apu(rdev); + if (ret) { + vmgmt_err(vdev, "Failed to bringup APU, ret %d", ret); + ret = -ENODEV; + goto queue_fini; + } + + rm_install_health_monitor(rdev); + + return rdev; +queue_fini: + rm_queue_fini(rdev); +err: + return ERR_PTR(ret); +} + +int vmgmt_rm_get_fw_id(struct rm_device *rdev, uuid_t *uuid) +{ + char str[UUID_STRING_LEN]; + ssize_t len = PAGE_SIZE; + char *buffer = NULL; + struct rm_cmd *cmd; + u8 i, j; + int ret; + + buffer = vmalloc(len); + if (!buffer) + return -ENOMEM; + + memset(buffer, 0, len);vzalloc()?
Yes, I will fix it.
+ + ret = rm_queue_create_cmd(rdev, RM_QUEUE_OP_GET_LOG_PAGE, &cmd); + if (ret) + return ret; + + ret = rm_queue_payload_init(cmd, RM_CMD_LOG_PAGE_FW_ID); + if (ret) + goto error; + + ret = rm_queue_send_cmd(cmd, RM_CMD_WAIT_CONFIG_TIMEOUT); + if (ret) + goto payload; + + ret = rm_queue_copy_response(cmd, buffer, len); + if (ret) + goto payload; + + /* parse uuid into a valid uuid string format */ + for (i = 0, j = 0; i < strlen(buffer); i++) { + str[j++] = buffer[i]; + if (j == 8 || j == 13 || j == 18 || j == 23) + str[j++] = '-'; + } + + uuid_parse(str, uuid); + vmgmt_dbg(rdev->vdev, "Interface uuid %pU", uuid); + + vfree(buffer); + + rm_queue_payload_fini(cmd); + rm_queue_destory_cmd(cmd); + + return 0; + +payload: + rm_queue_payload_fini(cmd); +error: + rm_queue_destory_cmd(cmd); + vfree(buffer); + return ret; +} diff --git a/drivers/fpga/amd/vmgmt-rm.h b/drivers/fpga/amd/vmgmt-rm.h new file mode 100644 index 000000000000..a74f93cefbe8 --- /dev/null +++ b/drivers/fpga/amd/vmgmt-rm.h @@ -0,0 +1,222 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Driver for Versal PCIe device + * + * Copyright (C) 2024 Advanced Micro Devices, Inc. All rights reserved. + */ + +#ifndef __VMGMT_RM_H +#define __VMGMT_RM_H + +#define RM_HDR_OFF 0x0 +#define RM_HDR_MAGIC_NUM 0x564D5230 +#define RM_QUEUE_HDR_MAGIC_NUM 0x5847513F +#define RM_PCI_IO_BAR_OFF 0x2010000 +#define RM_PCI_IO_SIZE 0x1000 +#define RM_PCI_SHMEM_BAR_OFF 0x8000000 +#define RM_PCI_SHMEM_SIZE 0x8000000 /* 128 MB */ +#define RM_PCI_SHMEM_HDR_SIZE 0x28 + +#define RM_QUEUE_HDR_MAGIC_NUM_OFF 0x0 +#define RM_IO_SQ_PIDX_OFF 0x0 +#define RM_IO_CQ_PIDX_OFF 0x100 + +#define RM_CMD_ID_MIN 1 +#define RM_CMD_ID_MAX (BIT(17) - 1) +#define RM_CMD_SQ_HDR_OPS_MSK GENMASK(15, 0) +#define RM_CMD_SQ_HDR_SIZE_MSK GENMASK(14, 0) +#define RM_CMD_SQ_SLOT_SIZE 512 +#define RM_CMD_CQ_SLOT_SIZE 16 +#define RM_CMD_CQ_BUFFER_SIZE (1024 * 1024) +#define RM_CMD_CQ_BUFFER_OFFSET 0x0 +#define RM_CMD_LOG_PAGE_TYPE_MASK GENMASK(15, 0) +#define RM_CMD_VMR_CONTROL_MSK GENMASK(10, 8) +#define RM_CMD_VMR_CONTROL_PS_MASK BIT(9) + +#define RM_CMD_WAIT_CONFIG_TIMEOUT msecs_to_jiffies(10 * 1000) +#define RM_CMD_WAIT_DOWNLOAD_TIMEOUT msecs_to_jiffies(300 * 1000) + +#define RM_COMPLETION_TIMER (HZ / 10) +#define RM_HEALTH_CHECK_TIMER (HZ) + +#define RM_INVALID_SLOT 0 + +enum rm_queue_opcode { + RM_QUEUE_OP_LOAD_XCLBIN = 0x0, + RM_QUEUE_OP_GET_LOG_PAGE = 0x8, + RM_QUEUE_OP_LOAD_FW = 0xA, + RM_QUEUE_OP_LOAD_APU_FW = 0xD, + RM_QUEUE_OP_VMR_CONTROL = 0xE, + RM_QUEUE_OP_IDENTIFY = 0x202, +}; + +struct rm_cmd_sq_hdr { + u16 opcode; + u16 msg_size; + u16 id; + u16 reserved; +} __packed; + +struct rm_cmd_cq_hdr { + u16 id; + u16 reserved; +} __packed; + +struct rm_cmd_sq_bin { + u64 address; + u32 size; + u32 reserved1; + u32 reserved2; + u32 reserved3; + u64 reserved4; +} __packed; + +struct rm_cmd_sq_log_page { + u64 address; + u32 size; + u32 reserved1; + u32 type; + u32 reserved2; +} __packed; + +struct rm_cmd_sq_ctrl { + u32 status; +} __packed; + +struct rm_cmd_sq_data { + union { + struct rm_cmd_sq_log_page page; + struct rm_cmd_sq_bin bin; + struct rm_cmd_sq_ctrl ctrl; + }; +} __packed; + +struct rm_cmd_cq_identify { + u16 major; + u16 minor; + u32 reserved; +} __packed; + +struct rm_cmd_cq_log_page { + u32 len; + u32 reserved; +} __packed; + +struct rm_cmd_cq_control { + u16 status; + u16 reserved1; + u32 reserved2; +} __packed; + +struct rm_cmd_cq_data { + union { + struct rm_cmd_cq_identify identify; + struct rm_cmd_cq_log_page page; + struct rm_cmd_cq_control ctrl; + u32 reserved[2]; + }; + u32 rcode; +} __packed; + +struct rm_cmd_sq_msg { + struct rm_cmd_sq_hdr hdr; + struct rm_cmd_sq_data data; +} __packed; + +struct rm_cmd_cq_msg { + struct rm_cmd_cq_hdr hdr; + struct rm_cmd_cq_data data; +} __packed; + +struct rm_cmd { + struct rm_device *rdev; + struct list_head list; + struct completion executed; + struct rm_cmd_sq_msg sq_msg; + struct rm_cmd_cq_msg cq_msg; + enum rm_queue_opcode opcode; + void *buffer; + ssize_t size; +}; + +enum rm_queue_type { + RM_QUEUE_SQ, + RM_QUEUE_CQ +}; + +enum rm_cmd_log_page_type { + RM_CMD_LOG_PAGE_AXI_TRIP_STATUS = 0x0, + RM_CMD_LOG_PAGE_FW_ID = 0xA, +}; + +struct rm_queue { + enum rm_queue_type type; + u32 pidx; + u32 cidx; + u32 offset; + u32 data_offset; + u32 data_size; + struct semaphore data_lock; +}; + +struct rm_queue_header { + u32 magic; + u32 version; + u32 size; + u32 sq_off; + u32 sq_slot_size; + u32 cq_off; + u32 sq_cidx; + u32 cq_cidx; +}; + +struct rm_header { + u32 magic; + u32 queue_base; + u32 queue_size; + u32 status_off; + u32 status_len; + u32 log_index; + u32 log_off; + u32 log_size; + u32 data_start; + u32 data_end; +}; + +struct rm_device { + struct vmgmt_device *vdev; + struct regmap *shmem_regmap; + struct regmap *io_regmap; + + struct rm_header rm_metadata; + u32 queue_buffer_start; + u32 queue_buffer_size; + u32 queue_base; + + /* Lock to queue access */ + struct mutex queue; + struct rm_queue sq; + struct rm_queue cq; + u32 queue_size; + + struct timer_list msg_timer; + struct work_struct msg_monitor; + struct timer_list health_timer; + struct work_struct health_monitor; + struct list_head submitted_cmds; + + int firewall_tripped; +}; + +int rm_queue_create_cmd(struct rm_device *rdev, enum rm_queue_opcode opcode, + struct rm_cmd **cmd_ptr); +void rm_queue_destory_cmd(struct rm_cmd *cmd); + +int rm_queue_data_init(struct rm_cmd *cmd, const char *buffer, ssize_t size); +void rm_queue_data_fini(struct rm_cmd *cmd); + +int rm_queue_copy_response(struct rm_cmd *cmd, void *buffer, ssize_t len); + +int rm_boot_apu(struct rm_device *rdev); + +#endif /* __VMGMT_RM_H */ diff --git a/drivers/fpga/amd/vmgmt.c b/drivers/fpga/amd/vmgmt.c index b72eff9e8bc0..198213a13c7d 100644 --- a/drivers/fpga/amd/vmgmt.c +++ b/drivers/fpga/amd/vmgmt.c @@ -21,6 +21,8 @@ #include "vmgmt.h" #include "vmgmt-comms.h" +#include "vmgmt-rm.h" +#include "vmgmt-rm-queue.h" #define DRV_NAME "amd-vmgmt" #define CLASS_NAME DRV_NAME @@ -43,6 +45,61 @@ static inline struct vmgmt_device *vmgmt_inode_to_vdev(struct inode *inode) return (struct vmgmt_device *)container_of(inode->i_cdev, struct vmgmt_device, cdev); } +static int vmgmt_fpga_write_init(struct fpga_manager *mgr, + struct fpga_image_info *info, const char *buf, + size_t count) +{ + struct fpga_device *fdev = mgr->priv; + struct fw_tnx *tnx = &fdev->fw; + int ret; + + ret = rm_queue_create_cmd(fdev->vdev->rdev, tnx->opcode, &tnx->cmd); + if (ret) { + fdev->state = FPGA_MGR_STATE_WRITE_INIT_ERR; + return ret; + } + + fdev->state = FPGA_MGR_STATE_WRITE_INIT; + return ret; +} + +static int vmgmt_fpga_write(struct fpga_manager *mgr, const char *buf, + size_t count) +{ + struct fpga_device *fdev = mgr->priv; + int ret; + + ret = rm_queue_data_init(fdev->fw.cmd, buf, count); + if (ret) { + fdev->state = FPGA_MGR_STATE_WRITE_ERR; + rm_queue_destory_cmd(fdev->fw.cmd); + return ret; + } + + fdev->state = FPGA_MGR_STATE_WRITE; + return ret; +} + +static int vmgmt_fpga_write_complete(struct fpga_manager *mgr, + struct fpga_image_info *info) +{ + struct fpga_device *fdev = mgr->priv; + int ret; + + ret = rm_queue_send_cmd(fdev->fw.cmd, RM_CMD_WAIT_DOWNLOAD_TIMEOUT); + if (ret) { + fdev->state = FPGA_MGR_STATE_WRITE_COMPLETE_ERR; + vmgmt_err(fdev->vdev, "Send cmd failed:%d, cid:%d", ret, fdev->fw.id); + } else { + fdev->state = FPGA_MGR_STATE_WRITE_COMPLETE; + } + + rm_queue_data_fini(fdev->fw.cmd); + rm_queue_destory_cmd(fdev->fw.cmd); + memset(&fdev->fw, 0, sizeof(fdev->fw)); + return ret; +} + static enum fpga_mgr_states vmgmt_fpga_state(struct fpga_manager *mgr) { struct fpga_device *fdev = mgr->priv; @@ -51,6 +108,9 @@ static enum fpga_mgr_states vmgmt_fpga_state(struct fpga_manager *mgr) } static const struct fpga_manager_ops vmgmt_fpga_ops = { + .write_init = vmgmt_fpga_write_init, + .write = vmgmt_fpga_write, + .write_complete = vmgmt_fpga_write_complete, .state = vmgmt_fpga_state, }; @@ -96,6 +156,13 @@ static struct fpga_device *vmgmt_fpga_init(struct vmgmt_device *vdev) return ERR_PTR(ret); } + ret = vmgmt_rm_get_fw_id(vdev->rdev, &vdev->intf_uuid); + if (ret) { + vmgmt_warn(vdev, "Failed to get interface uuid"); + ret = -EINVAL; + goto unregister_fpga_mgr; + } + /* create fgpa bridge, region for the base shell */ fdev->bridge = fpga_bridge_register(dev, "AMD Versal FPGA Bridge", &vmgmt_br_ops, fdev); @@ -132,6 +199,149 @@ static struct fpga_device *vmgmt_fpga_init(struct vmgmt_device *vdev) return ERR_PTR(ret); } +static int vmgmt_region_program(struct fpga_region *region, const void *data) +{ + struct fpga_device *fdev = region->priv; + struct vmgmt_device *vdev = fdev->vdev; + const struct axlf *xclbin = data; + struct fpga_image_info *info; + int ret; + + info = fpga_image_info_alloc(&vdev->pdev->dev); + if (!info) + return -ENOMEM; + + region->info = info; + + info->flags |= FPGA_MGR_PARTIAL_RECONFIG; + info->count = xclbin->header.length; + info->buf = (char *)xclbin; + + ret = fpga_region_program_fpga(region); + if (ret) { + vmgmt_err(vdev, "Programming xclbin failed: %d", ret); + goto exit; + } + + /* free bridges to allow reprogram */ + if (region->get_bridges) + fpga_bridges_put(®ion->bridge_list); + +exit: + fpga_image_info_free(info); + return ret; +} + +static int vmgmt_fpga_region_match(struct device *dev, const void *data) +{ + const struct vmgmt_fpga_region *arg = data; + const struct fpga_region *match_region; + struct fpga_device *fdev = arg->fdev; + uuid_t compat_uuid; + + if (dev->parent != &fdev->vdev->pdev->dev) + return false; + + match_region = to_fpga_region(dev); + + import_uuid(&compat_uuid, (const char *)match_region->compat_id); + if (uuid_equal(&compat_uuid, arg->uuid)) { + vmgmt_dbg(fdev->vdev, "Region match found"); + return true; + } + + vmgmt_err(fdev->vdev, "download uuid %pUb is not the same as device uuid %pUb", + arg->uuid, &compat_uuid); + return false; +} + +static long vmgmt_ioctl(struct file *filep, unsigned int cmd, unsigned long arg) +{ + struct vmgmt_device *vdev = (struct vmgmt_device *)filep->private_data; + struct vmgmt_fpga_region reg = { 0 }; + struct fpga_region *region = NULL; + struct axlf *axlf = NULL; + void *data = NULL; + size_t size = 0; + int ret = 0; + + axlf = vmalloc(sizeof(*axlf)); + if (!axlf) + return -ENOMEM; + + ret = copy_from_user((void *)axlf, (void *)arg, sizeof(*axlf)); + if (ret) { + vmgmt_err(vdev, "Failed to copy axlf: %d", ret); + ret = -EFAULT; + goto exit; + } + + ret = memcmp(axlf->magic, VERSAL_XCLBIN_MAGIC_ID, + sizeof(VERSAL_XCLBIN_MAGIC_ID)); + if (ret) { + vmgmt_err(vdev, "unknown axlf magic %s", axlf->magic); + ret = -EINVAL; + goto exit; + } + + /* axlf should never be over 1G and less than size of struct axlf */ + size = axlf->header.length; + if (size < sizeof(struct axlf) || size > 1024 * 1024 * 1024) { + vmgmt_err(vdev, "axlf length %zu is invalid", size); + ret = -EINVAL; + goto exit; + } + + data = vmalloc(size); + if (!data) { + ret = -ENOMEM; + goto exit; + } + + ret = copy_from_user((void *)data, (void *)arg, size); + if (ret) { + vmgmt_err(vdev, "Failed to copy data: %d", ret); + ret = -EFAULT; + goto exit; + } + + switch (cmd) { + case VERSAL_MGMT_LOAD_XCLBIN_IOCTL: + vdev->fdev->fw.opcode = RM_QUEUE_OP_LOAD_XCLBIN; + break; + default: + vmgmt_err(vdev, "Invalid IOCTL command: %d", cmd); + ret = -EINVAL; + goto exit; + } + + reg.uuid = &axlf->header.rom_uuid; + reg.fdev = vdev->fdev; + + region = fpga_region_class_find(NULL, ®, vmgmt_fpga_region_match); + if (!region) { + vmgmt_err(vdev, "Failed to find compatible region"); + ret = -ENOENT; + goto exit; + } + + ret = vmgmt_region_program(region, data); + if (ret) { + vmgmt_err(vdev, "Failed to program region"); + goto exit; + } + + vmgmt_info(vdev, "Downloaded axlf %pUb of size %zu Bytes", + &axlf->header.uuid, size); + uuid_copy(&vdev->xclbin_uuid, &axlf->header.uuid); + +exit: + vfree(data); + vfree(axlf); + + return ret; +} + static int vmgmt_open(struct inode *inode, struct file *filep) { struct vmgmt_device *vdev = vmgmt_inode_to_vdev(inode); @@ -155,6 +365,7 @@ static const struct file_operations vmgmt_fops = { .owner = THIS_MODULE, .open = vmgmt_open, .release = vmgmt_release, + .unlocked_ioctl = vmgmt_ioctl, }; static void vmgmt_chrdev_destroy(struct vmgmt_device *vdev) @@ -201,6 +412,69 @@ static int vmgmt_chrdev_create(struct vmgmt_device *vdev) return 0; } +static enum fw_upload_err vmgmt_fw_prepare(struct fw_upload *fw_upload, + const u8 *data, u32 size) +{ + struct firmware_device *fwdev = fw_upload->dd_handle; + struct axlf *xsabin = (struct axlf *)data; + int ret; + + ret = memcmp(xsabin->magic, VERSAL_XCLBIN_MAGIC_ID, + sizeof(VERSAL_XCLBIN_MAGIC_ID)); + if (ret) { + vmgmt_err(fwdev->vdev, "Invalid device firmware"); + return FW_UPLOAD_ERR_INVALID_SIZE; + } + + /* Firmware size should never be over 1G and less than size of struct axlf */ + if (!size || size != xsabin->header.length || size < sizeof(*xsabin) || + size > 1024 * 1024 * 1024) { + vmgmt_err(fwdev->vdev, "Invalid device firmware size"); + return FW_UPLOAD_ERR_INVALID_SIZE; + } + + ret = rm_queue_create_cmd(fwdev->vdev->rdev, RM_QUEUE_OP_LOAD_FW, + &fwdev->cmd); + if (ret) + return FW_UPLOAD_ERR_RW_ERROR; + + uuid_copy(&fwdev->uuid, &xsabin->header.uuid); + return FW_UPLOAD_ERR_NONE; +} + +static enum fw_upload_err vmgmt_fw_write(struct fw_upload *fw_upload, + const u8 *data, u32 offset, u32 size, + u32 *written) +{ + struct firmware_device *fwdev = fw_upload->dd_handle; + int ret; + + ret = rm_queue_data_init(fwdev->cmd, data, size); + if (ret) + return FW_UPLOAD_ERR_RW_ERROR; + + *written = size; + return FW_UPLOAD_ERR_NONE; +} + +static enum fw_upload_err vmgmt_fw_poll_complete(struct fw_upload *fw_upload) +{ + struct firmware_device *fwdev = fw_upload->dd_handle; + int ret; + + vmgmt_info(fwdev->vdev, "Programming device firmware: %pUb", &fwdev->uuid); + + ret = rm_queue_send_cmd(fwdev->cmd, RM_CMD_WAIT_DOWNLOAD_TIMEOUT); + if (ret) { + vmgmt_err(fwdev->vdev, "Send cmd failedi:%d, cid %d", ret, fwdev->id); + return FW_UPLOAD_ERR_HW_ERROR; + } +Basically I didn't see any difference on HW operations between FPGA reprogram & firmware loading. So why use different SW interfaces? My idea is FPGA reprogramming is not just for image loading, but also device re-enumeration. As I mentioned before, programing the fpga region without notifying the impacted driver makes kernel crash. I see there is an effort to introduce a unified FPGA reprograming interface that should address the concerns. I think we should stop adding vendor interfaces and make effort on the unified one. https://lore.kernel.org/linux-fpga/20240726063819.2274324-1-nava.kishore.manne@xxxxxxx/
This Versal card has 3 data images that we need to download.1) The base FW - that enables the hardware, only take effect with a cold reboot.
2) The APU FW - the additional FW that is compatible with the base FW. 3) The PL Data - the (FPGA Bitstream Binary) can change the hardware PL.Our original design was using ioctl to download No.1, and using the request_firmware to download the No.2. Both No.1 and No.2 will not be dynamically re-programable. We took the existing "intel-m10-bmc-sec-update.c" in the fpga as an example. This was why we converted the No.1 ioctl to fw_upload.
The No.3 will be initiated by the user PF driver only. Then, the mgmt PF driver will use request_firmware to load local saved data and then download it onto the FPGA. Thus, in this case, we don't need sysfs or ioctl. I will add those code in next patch.
Thus, we don't have to use new IOCTLs to load images for No.1, No.2 and No.3. The existing request_firmware API works for all.
I also briefly chatted with "nava.kishore.manne@xxxxxxx". He is going to add ioctl and leveraging fpga_region_config_info for OF driver for now.
+ vmgmt_info(fwdev->vdev, "Successfully programmed device firmware: %pUb", + &fwdev->uuid); + return FW_UPLOAD_ERR_NONE; +} + static void vmgmt_fw_cancel(struct fw_upload *fw_upload) { struct firmware_device *fwdev = fw_upload->dd_handle; @@ -208,8 +482,26 @@ static void vmgmt_fw_cancel(struct fw_upload *fw_upload) vmgmt_warn(fwdev->vdev, "canceled");Nothing to do?
The rm_queue_withdraw_cmd should be called. I will fix this.
I'll stop here, please better organize the patches next time submitting to make reviewers easier, even if you know it is an RFC.
Thanks for taking your time reviewing all those patches. I will reorganize the patches.
Thanks, Yilun
